The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill demonstrates the use of load_learner('classifier_model.pkl'), which utilizes Python's pickle module. Deserializing data from a potentially untrusted .pkl file is a known security risk that can lead to arbitrary code execution if the file has been tampered with or originates from an unverified source.
[INDIRECT_PROMPT_INJECTION]: The skill provides patterns for processing external data, including CSV files (pd.read_csv) and image datasets (ImageDataLoaders). This introduces an attack surface for indirect prompt injection, where malicious instructions embedded in the data could attempt to influence the agent's behavior during runtime.
Ingestion points: External files are loaded via pd.read_csv('customers.csv') and untar_data(URLs.PETS) in SKILL.md.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the provided snippets.
Capability inventory: The skill possesses capabilities for file system access (read/write) and model training, though it does not explicitly invoke shell commands.
Sanitization: There is no evidence of sanitization or validation of the content within the processed data files.
[EXTERNAL_DOWNLOADS]: The skill uses the untar_data function to fetch the PETS dataset from fast.ai's official resources. This is a standard utility within the library for retrieving well-known datasets.

fastai